One of the most massive AWS incidents transpired on October 20th. The long story short is that the DNS for DynamoDB was impacted for us-east-1, which created a health event for the entire region. It’s the worst incident we’ve seen in a decade. authress shares with us how they maintain a high SLI to match our SLA reliability commitment even when the infrastructure and services we use don’t.

Part 12 of “How to Scale Your Model”, an free ebook that aim to demystify the science of scaling language models: how TPUs (and GPUs) work and how they communicate with each other, how LLMs run on real hardware, and how to parallelize your models during training and inference so they run efficiently at massive scale.

Memory maps are a modern Unix mechanism where you can take a file and make it part of the virtual memory. In Unix context, modern means that it was introduced in the 1980s or later. You have a file, containing data, you mmap it and you’ll get a pointer to where this resides. Now, instead of seeking and reading, you just read from this pointer, adjusting the offset to get to the right data.

The original motivation for the creation of Bloom filters is efficient set membership, using a probabilistic approach to significantly reduce the time and space required to reject items that are not members in a certain set.

It stands for “Authenticated Encryption with Associated Data” but if you are struggling to remember what is it. This post sumarize nicely.

There are many SQL JOINs guides and tutorials, but this one takes a different approach. We try to avoid misleading wording and imagery, and we structure the material in a different way. The goal of this article is to clarify your mental model

The secret to unlocking performance gains often lies not just in what you ask in a query, but in how Postgres finds the answer. The Postgres EXPLAIN system is great for understanding how data is being queried. One of secretes to reading EXPLAIN plans is understanding the type of scan done to retrieve the data. The scan type can be the difference between a lightning-fast response or a slow quey

Addy Osmani is a legendary engineer. In his own words: “When I joined Google ~14 years ago, I thought the job was about writing great code. I was partly right. But the longer I’ve stayed, the more I’ve realized that the engineers who thrive aren’t necessarily the best programmers - they’re the ones who’ve figured out how to navigate everything around the code: the people, the politics, the alignment, the ambiguity.”

WebPerformance Report was designed around that principle. Although its underlying data comes from automated performance tests, the report itself is intentionally high-level. It is structured to help anyone, developers, designers, marketers, and executives, understand what happened, why it matters, and where the opportunities lie. This article walks through the anatomy of a Web Performance Report, explaining each section, why it exists, and what decisions it enables.

To implement accessibility, Chrome has this pipeline DOM mutation -> AXEvent -> Accessibility tree -> OS accessibility API -> Screen reader. But here’s the strange part: Chrome doesn’t have one accessibility tree, it has two. This article traces the full lifecycle of an accessibility event. From the moment the DOM changes, through Blink’s internal systems, through its internal accessibility tree, and finally back out into the BrowserAccessibility tree that screen readers query.